[bind10-dev] resolver performance research, take 2

[JINMEI Tatuya / 神明達哉]

At Tue, 17 Jul 2012 09:45:51 +0200,
Michal 'vorner' Vaner <michal.vaner at nic.cz> wrote:

> > - We might even introduce an optimization skipping any TTL adjustment
> >   in that cache
> 
> As I'm looking at the „I don't need to re-render for 1 second, as
> the TTL does not change“ thing, I think it might be worth to render
> the answer with a lower TTL than there actually is and keep it
> longer (eg. if the answer would have 300 as TTL, render it with 290
> and keep it rendered for 10s instead of one).

That's another option.  I mentioned even more radical idea of
consistently using a TTL of 0 (until the real TTL expires) in the
previous discussion, but...

> Currently, does any stub resolver keep a cache or something?

MacOS X (apparently) does.  I guess it's the same for iOS.  I don't
know about Windows, which probably matters most in this context.  So
reducing it too aggressively is probably a bad idea.

---
JINMEI, Tatuya
Internet Systems Consortium, Inc.